Right for the Right Reasons

Transcription

1 Right for the Right Reasons A final rejoinder on the Stern Review Simon Dietz, Dennis Anderson, Nicholas Stern, Chris Taylor & Dimitri Zenghelis 1. Introduction In this, our final rejoinder on the debate about the Stern Review that has been published by World Economics, 1 we respond to comments in this issue by Robert Carter et al., by David Henderson, and by Richard Tol and Gary Yohe. 2 Carter et al. continue to argue against a growing body of scientific evidence and a growing consensus on that same evidence: the climate system is now warming significantly; this warming is more likely than not to continue and could be rapid; human activities are the major cause of it; potentially very large risks are involved; hence it is an immediate priority for public policy to pursue both greenhouse gas (GHG) emission reductions (mitigation) and adaptation. The source of their critique is, first, a distinctly partisan, and increasingly untenable, position on the broad range of available scientific evidence and, second, a mistrust of the international consensus-building exercise centred on the Intergovernmental Panel on Climate Change (IPCC). Henderson is also largely preoccupied with the Simon Dietz is Lecturer in Environmental Policy at the London School of Economics and Political Science, and a Visiting Fellow at the Tyndall Centre for Climate Change Research. Dennis Anderson is Emeritus Professor of Energy and Environmental Studies at Imperial College London. Nicholas Stern was Head of the Stern Review of the Economics of Climate Change, and Head of the UK s Government Economic Service. In June 2007, he took up a new appointment as the first holder of the IG Patel Chair in Economics and Government at the London School of Economics and Political Science. Chris Taylor is an Economic Advisor to the Stern Review of the Economics of Climate Change, HM Treasury. Dimitri Zenghelis is Senior Economist on the Stern Review, HM Treasury. The authors would like to acknowledge helpful comments from Siobhan Peters and Leonard Smith. 1 See the various contributions to issue four of volume seven and issue one of volume eight. 2 Robert M. Carter, C. R. de Freitas, Indur M. Goklany, David Holland & Richard S. Lindzen, Climate Science and the Stern Review, pp ; David Henderson, Governments and Climate Change Issues: The case for rethinking, pp ; Richard S. J. Tol & Gary W. Yohe, A Stern Reply to the Reply to the Review of the Stern Review, pp WORLD ECONOMICS Vol. 8 No. 2 April June

2 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis latter, procedural issues. Tol and Yohe focus on economic arguments. It is important to remind readers that, in sharp contrast to Carter et al., Tol and Yohe support our interpretation of the current scientific evidence. Their critique is rather narrower in focus and concerns the way in which abatement costs were calculated in the supporting work carried out by one of us (Anderson, 2006). It rests on basic confusions and misconceptions, many of which were explained in our previous contributions (Anderson, 2007; Dietz et al., 2007). However, we think readers of World Economics will be more interested in a broader reflection. How would we, following the debate of the last eight months, assess the approach, policies and arguments set out in the Review? In our view, our analyses and policy proposals, and the arguments in support, are sound and have stood up well to scrutiny. In other words, we were right and for the right reasons. In summary, what are those positions? First, that the cost of strong and urgent action is much less than the cost of delayed or timid action. Second, that policy should give priority to a time path of emissions that can lead to stabilisation of the atmospheric stock of GHGs in the range of 450 to 550 parts per million carbon dioxide equivalent (ppm CO 2 e), and that this requires emissions to peak within twenty years and to fall by at least 30% by Third, that market mechanisms will be a crucial element in guiding emission reductions in a costeffective way, together with appropriate regulation and standards, including for energy efficiency. Fourth, strong action on research, development and the deployment of new techniques will be required. Fifth, action on deforestation is urgent and very cost-effective. Sixth, adaptation will be of great importance for both developing and developed countries, with the former hit earliest and hardest. Seventh, any global deal should embody the above principles, be constructed in a way that takes strong account of equity between rich and poor nations, both on mitigation and adaptation, and promote an understanding of the risks to economic competitiveness and the opportunities, where early action is taken by individual states, regions and companies. Central to many critiques of the Review (e.g. Dasgupta, 2006; Mendelsohn, 2006; Neumayer, forthcoming; Nordhaus, 2006; Tol and Yohe, 2006 and 2007; Weitzman, forthcoming) is a fundamental misunderstanding of the 3 By up to 70% for stabilisation at 450 ppm CO 2 e. 230 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

3 Right for the Right Reasons: A final rejoinder on the Stern Review role of formal, highly aggregated economic modelling in evaluating a policy issue characterised by the very long run, by profound ethical considerations, great uncertainty, market imperfections, limited policy instruments and a requirement for international collaboration. It is this misunderstanding, which sometimes fosters the claim that the Review is in fact right, but for the wrong reasons. Disproportionate attention is thus drawn to the small part of the Review devoted to such formal modelling, especially of the impacts of climate change, so that critics fail to consider the full range of evidence we presented (e.g. Weitzman, forthcoming). At the same time, excessive focus on the formal modelling implies that real-world policy prescriptions become hostage to choosing the values of highly aggregated and simplified parameters within the models, together with the relationships between them (e.g. Nordhaus, 2006). Formal models can and should play an important role in the systematic and transparent exploration of assumptions and value judgements, and how they affect the scale and structure of policy. But it is misleading and dangerous to base policy primarily on the results of such models. That is why just one of the 13 chapters comprising the first half of the Review (the second half was on the details of policy instruments) was devoted to integrated assessment models of the monetary cost of climate change, 4 why we were so clear on their limitations, and why our emphasis was on the more detailed regional effects on the many relevant dimensions of human welfare. Nevertheless, we have argued strongly and in our view convincingly (see the Postscript to the Review 5 and Dietz et al., 2007) that, even within the confines of formal economic modelling, the concerns raised by a small group of commentators do not overturn our basic conclusion that the cost of action is much less than the cost of inaction. Our critics here fall short by failing to simultaneously afford the necessary importance to issues of risk and ethics. The case for strong and urgent action set out in the Review is based, first, on the severe risks that the science now identifies (together with the additional uncertainties 6 that it points to but that are difficult to quantify) and, second, on the ethics of the responsibilities of existing 4 Chapter 10 analysed macroeconomic models of the costs of emission reductions. 5 References to the Review are to The Economics of Climate Change: The Stern Review, 2007, published in January by Cambridge University Press. The Review was published on the web, at at the end of October 2006 and a Postscript added around a month later (and included in the CUP book). 6 Where we distinguish between risk and uncertainty, we adopt the Knightian approach to the latter concept: i.e. it corresponds to circumstances where we are not in a position to attach probabilities to uncertain events. WORLD ECONOMICS Vol. 8 No. 2 April June

4 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis generations in relation to succeeding generations. It is these two things that are crucial: risk and ethics. Different commentators may vary in their emphasis, but it is the two together that are crucial. Jettison either one and you will have a much reduced programme for action and if you judge risks to be small and attach little significance to future generations you will not regard global warming as a problem. It is surprising that the earlier economic literature on climate change did not give risk and ethics the attention they so clearly deserve, and it is because we chose to make them central and explicit that we think we were right for the right reasons. The implications of following a slow policy ramp, with meagre emission reductions over the next quarter of a century, would, on our reading of the evidence, be very risky indeed, yet this is implicit in the arguments of, for example, Nordhaus (2006) and Tol and Yohe (2006 and 2007). They do not seem ready to acknowledge the riskiness of the paths they suggest. Our conclusion, that global efforts should be directed at stabilising GHG concentrations in the range of 450 to 550 ppm CO 2 e, is robust to a variety of considerations (Dietz et al., 2007). If our critics are prepared to allow GHG concentrations to rise to 650 ppm CO 2 e, 750, or beyond, with significant associated risks of eventual temperature increases in excess of 5 C above pre-industrial levels (a transformation in global climate, taking it way beyond human experience, and making radical relocations of populations likely), then they should say so. They would have to explicitly argue that the risks brought about by such temperature changes are small, or confidently state that we can adapt to the huge changes in environment brought about by 5 C warming and that we can do so cheaply, or that we simply do not care about these risks and costs, because they are a problem for future generations. 2. The scientific critique Recent observations of climatic changes, together with improvements to the historical record from multiple data sources, have put beyond any reasonable doubt the conclusion that the climate system is warming, and significantly so (e.g. table 3.2 of Trenberth et al., 2007). The conclusions of Carter et al. (2007) are untenable given this growing body of evidence, because they continue to rely on earlier research, notably Lindzen (1990), which were important contributions to the science at the time, but which 232 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

5 Right for the Right Reasons: A final rejoinder on the Stern Review are now remarkably inconsistent with observed warming over the past two decades or so. Eleven of the twelve warmest years on record since 1850 have been between 1995 and 2006, after Lindzen s paper (Solomon et al., 2007). 7 Furthermore, as Mitchell et al. (2007) explained, this warming cannot be attributed to natural forcing alone: we can be very confident 8 that most of this warming is due to GHG emissions from human activities. This conclusion is drawn on the basis of running numerical atmosphere/ ocean General Circulation Models (GCMs) against the observational record. Only with GCMs can the various sources of warming, natural and from human activities, be considered separately and in combination. Only when GHG emissions from human activities are included can the observed increases in temperature over the 20th century be reproduced. In return, Carter et al. (2007) attempt to cast doubt on the value of GCMs. They seek to characterise climate modelling as fitting a curve to the observational data, so that a good fit between a GCM and the observational record apparently proves nothing at all. We certainly do not have the evidence to directly validate exercises in modelling future states of the climate system over long time-scales. Climate prediction is an exercise in extrapolation to states of the world that have, to all intents and purposes, never been experienced before. But modellers do not fit curves in the way Carter et al. suggest. On the contrary, our physical understanding of the key processes represented within GCMs constrains their structure strongly (see also Mitchell et al., 2007). This is not model and parameter manipulation to achieve a perfect fit. Indeed, it is important to note that, despite several decades of developing models and a million degrees of freedom with which to obtain explanatory power, GCM simulations are far from an adequate fit to the observational record, on its many dimensions. While models may replicate changes in global mean temperature quite well (Rahmstorf et al., 2007), they continue to represent other processes, more directly relevant to predicting the future impacts of climate change, either badly (e.g. hurricanes), or not at all (e.g. many characteristics of the El Niño Southern Oscillation: for a general discussion see Stainforth et al., forthcoming). 7 Carter et al. attempt to suppress this evidence by benchmarking recent changes in global mean temperature against the exceptionally warm year % confident according to the Summary for Policymakers of Working Group I of the IPCC s Fourth Assessment Report (IPCC, 2007). This is an expression of confidence based on expert judgement. WORLD ECONOMICS Vol. 8 No. 2 April June

6 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis Thus risk and uncertainty are, and are likely to remain, at the heart of the science and in turn the economics of climate change, and this was indeed a message that the Review repeatedly emphasised. One needs only to look, for example, at the confidence intervals around the estimates of the monetary cost of climate change in Figure 6.5, or to those around the eventual temperature change resulting from various stabilisation levels in Figure 13.4 (both in Stern, 2007). While these confidence intervals recognise that, on the basis of the most recent evidence (see e.g. Box 10.2 of IPCC, 2007), there is a small chance that business-as-usual GHG emissions could result in negligible warming, there is also a small chance that such increases could warm the Earth by significantly more than 5 C on average. It is precisely because the analysis of the Review entertains all these possibilities that the burden of proof lies, as Mitchell et al. (2007) propose, with the climate sceptics to demonstrate the basis of their confidence that warming caused by human sources of GHG emissions is small and there is trivial risk of substantial damage. At the core of the critique offered by Carter et al. (2007) and echoed by Henderson (2007) is a peculiar contradiction. While current syntheses of the evidence, of the type provided by IPCC and the Review, are charged with paying insufficient attention to uncertainty (an accusation that cannot in any case be squared with the analysis and presentation), climate sceptics immediately proceed to ignore all evidence to suggest that there may be a risk of rapid warming as a response to GHG emissions, making categorical assertions that there is, essentially, negligible risk. The fundamental question that anyone considering the debate in the pages of this journal needs to ask is, what decisions to make in the face of such risk and uncertainty? The central tenets of the economics of risk, including recent extensions to uncertainty (explained in Chapter 2 of the Review), emphasise the importance of reducing the risks of severe climate impacts. If that is accepted, then as we set out in the Review and in our previous paper (Dietz et al., 2007), the flow-stock mechanics driving the accumulation of GHGs in the atmosphere mean that delay in cutting back emissions is costly. Stocks are very difficult to reduce and thus the great uncertainties about future climate change are a reason to act, not to wait and see. In reality, of course, we do not face a once-and-forever decision. Rather we expect to revise policy on emission abatement as new information on both the costs and benefits of mitigation comes to light. 234 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

7 Right for the Right Reasons: A final rejoinder on the Stern Review Nevertheless, the prospect of learning does not support a wait-and-see strategy either (e.g. O Neill et al., 2006; Yohe et al., 2004). Indeed, it is very plausible that some learning will increase uncertainty around key climate parameters, as we improve our understanding of the way in which models represent reality (Stainforth et al., forthcoming). The atmospheric stock of GHGs is currently around 430 ppm CO 2 e and the rate of addition to that stock is around 2.5 ppm CO 2 e per annum and rising quickly. Together, these imply that delayed action would probably increase the stock to above 500 ppm in 25 years, making it very difficult to keep below 550 ppm. As we argued in the Review, 550 ppm CO 2 e is itself a risky place to be, with around a 50% probability that temperature increases will eventually exceed 3 C relative to pre-industrial times. Put simply, action, even in the unlikely event that the science is wrong, will give us some useful new technologies and investments. Timid action or inaction will, in the likely event that the science is right, put us in a very dangerous position, from which it will be extremely difficult to extricate ourselves. The economics of risk clearly points to strong action. Carter et al. (2007) and Henderson (2007) are often preoccupied with procedural concerns about the conduct of the IPCC and about standards of peer review and data disclosure. We believe these concerns have been blown to a proportion far beyond their actual importance to the evaluation of climate policy and to policy-making. As we noted in our previous contribution (Dietz et al., 2007), the IPCC Assessment Reports are the most important synthesis of scientific (and other) evidence on climate change. The process is complicated and may not be perfect so that Henderson s suggestions on strengthening the process deserve consideration but the outputs are carefully compiled and presented, and are extraordinarily comprehensive. The wide variety of evidence presented, on each and every side of the ledger, helps to build confidence. And, notwithstanding the current limits which modelling places on climate prediction, the diligence displayed by the IPCC in compiling the evidence is obvious. In a recent paper, Rahmstorf et al. (2007) show that IPCC projections of climatic changes for the period 1990 to 2005, based on GCMs, track the actual observational record very closely for atmospheric CO 2 concentrations and global mean temperature. 9 If anything, the observational record actually 9 These climate models were developed over many years prior to the period of comparison, and they are, as we have emphasised, constrained in important ways to represent basic physical processes rather than simply fitted to the data. WORLD ECONOMICS Vol. 8 No. 2 April June

8 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis indicates that the climate system, in particular sea-level, is responding more quickly to GHG emissions than the IPCC projections would have predicted. There is no evidence to suggest the IPCC process distorts the presentation of evidence in such a way as to exaggerate the risks of future climate change. 3. Abatement costs We respond here and in the Appendix to Tol and Yohe (2007) on the costs of emission abatement. They put forward a curve describing the global marginal costs of abatement, in which the marginal costs rise with the square of the amount of abatement at any point in time, but the curve as a whole shifts downwards at an exponential rate with exogenous technical progress, which is assumed to apply to all technologies equally. From this they deduce that, for our estimates to be reconciled with their view, our estimates of costs have to decline by factors of to over a 45-year period. This is, of course, an absurd exaggeration, which stems from an over-simplified indeed a flawed view of how innovation affects costs. Mitigation technologies such as fossil fuels with carbon capture and storage, nuclear power, solar energy, wind power, biofuels, hydrogen derived from carbon-neutral resources, low-carbon vehicles and several others are all at very different stages of development and show different rates of technical progress. Furthermore, technical progress does not appear exogenously with the passage of time, as Tol and Yohe assume, but endogenously with investment in R&D, demonstration and deployment. Although constraints will be encountered in some cases, acting to increase costs (the land requirements of biofuels is an obvious case in point, the need for storage technologies is another), ways round these constraints through substitution and technical progress can be anticipated, and indeed are already the subject of much scientific and engineering research in the universities, national research laboratories and industry. Progress will be uneven, but in general, the more we abate pollution the better we should become at it. The evidence on learning curves, for example for electricity generation, was set out in Chapter 9 of the Review. Moreover, the wide variety and scope of mitigation options across sectors and technologies is what allows expensive constraints in any one sector to be avoided. The ability to spread the burden of abatement across so many 236 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

9 Right for the Right Reasons: A final rejoinder on the Stern Review sectors is precisely what limits the total costs of stabilisation and hence it is essential that coordination and flexibility be achieved in global efforts to reduce emissions. Different approaches provide different insights and that is why the Review chose to draw on accessible, bottom-up estimates of costs based on an analysis of technological options (Chapter 9 and Anderson, 2006) and to complement this with the findings of the macroeconomic models that we thought captured behavioural effects well, while not misrepresenting (as Tol and Yohe have done) the importance of technical progress on pollution abatement (Chapter 10). A glance at Table 2 in the Appendix to this paper shows that our assumptions about technical progress are modest and also differ between technologies over the 45-year period; the assumed changes in costs are measured in percentage shifts, not by the astronomical orders of magnitude Tol and Yohe assert. These assumptions, which are also modest in relation to historical norms for the industry, were fully reported in the Review (Chapter 9) and the background paper by Anderson (2006) and are easy to compare with those of many other sources. Indeed, that was our intention: to set out the estimates and assumptions in ways that readers of the Review may easily check for themselves. We have re-tabulated them once more such that the readers may indeed form their own judgements. 4. Ethics and risk 4.1. A broad range of evidence Many critiques of the Stern Review are wide of the mark, because they fail either to recognise or to acknowledge the broad range of evidence that supported our conclusions. To read these critiques is to form the mistaken impression that formal economic modelling was the mainstay of the Review. Tol and Yohe (2007) appear to be particularly confused as to the role of our formal modelling, a confusion they are keen to foster through selective quotation from our Executive Summary. The Review s assessment was built on three lines of investigation. The simplest and most important comparison to be made is between our disaggregated analysis of the physical impacts of climate change on multiple metrics (e.g. water and food availability, health and infrastructure: Chapters 1, 3, 4 and 5), and WORLD ECONOMICS Vol. 8 No. 2 April June

10 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis bottom-up estimates of the costs of specific mitigation strategies, based on different portfolios of technologies (Chapter 9 and Anderson, 2006). As we re-emphasised in our previous paper (Dietz et al., 2007), the key question that policy makers should ask is whether paying an insurance premium of around 1% of GDP over much of this century is worthwhile to reduce the risks and uncertainties described. This question is central, because it presents the basic policy problem as simply and transparently as possible, thereby avoiding the process of aggregation of risks and uncertainties across all nations and dimensions, a process for which data are extremely thin and which ignores or suppresses so much of what is important. Nevertheless, formal economic models are useful for exploring particular, stylised aspects of the problem, such as the role of attitudes to intergenerational equity and risk in estimating the cost of climate change, and the role of behavioural changes in the economy as a whole in determining the cost of mitigation. Thus the second line of investigation in the Review compared the results of integrated assessment models of the cost of future climate change (impacts and adaptation: Chapter 6), with macro-economic models of the cost of mitigation (Chapter 10). The third approach, in Chapter 13 of the Review, set out an informal price-based approach, comparing the expected marginal costs of shifting from one path of emission reductions to another (e.g. moving from a stabilisation target of 650 to 550 ppm CO 2 e), with the expected marginal benefits of doing so. 10 In thinking about the role of formal models in climate change, arguments for trade liberalisation provide a useful analogy. These trade arguments are based on a range of perspectives, from detailed studies of experiences with different trade policies in different times and contexts, to basic conceptual notions, in particular comparative advantage. On top of this, there are a number of computable general equilibrium (CGE) models of trade, which attempt to directly estimate gains from liberalisation. Most economists are much more convinced by the former range of arguments than by the CGEs, because the latter leave out too much that is crucial and are often rather sensitive to assumptions concerning what they do include. Furthermore, in thinking about the analogy, we should recognise that CGE models of trade are probably much fitter for the purpose of 10 The three approaches are, of course, logically related in a formal sense, but they represent different perspectives on the problem. 238 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

11 Right for the Right Reasons: A final rejoinder on the Stern Review capturing trade effects than integrated assessment models are for capturing the economic effects of climate change. But even within the narrow confines of formal economic modelling, the concerns raised by many commentators do not overturn our basic conclusion that the costs of timid or delayed action are very high and substantially exceed the costs of timely, measured and well-planned emission reductions. Ultimately, the debate turns on the fundamental structure of the problem defining climate-change policy: on the one hand, ethics, both in terms of the ethical status of future generations and the distribution of income and wealth, and on the other, risk Ethics and discounting in formal economic modelling In formal economic modelling, the ethical discussion has been primarily focused on attitudes to inequality (via the elasticity of the social marginal utility of consumption, η) and the weight given to future generations (via the rate of pure time preference, δ). This is, of course, already a very narrow view of ethics, which for example omits notions of rights and responsibilities between and within generations. Nevertheless, in these highly aggregated models, ethical considerations usually boil down to these very simplistic concerns. But there is another problem that cannot be made to disappear even in these narrow models. That is, the consequences of climate change can be very large for the world as a whole. In a context where we must consider major non-marginal changes, the marginal approach of standard investment appraisal, which depends on seeing projects as small perturbations around a given path of economic growth, is inadequate. The risks that climate change poses could significantly reduce economic growth in the future, so we cannot simply take growth as given. Instead, we have to use the same economic and ethical principles, but go back to the overall objective of social welfare and the overall model structure that underpins the special case of the marginal approach. We must be clear on this: the transformation of the planet at issue here cannot be considered marginal. It is therefore a basic mistake to start the ethical analysis with an initial view of appropriate discount rates (e.g. Nordhaus, 2006; Weitzman, forthcoming). There are very difficult issues in starting a numerical discussion from market discount rates even within standard, medium-term cost benefit analysis of marginal changes. In other words, even in this context, it is a WORLD ECONOMICS Vol. 8 No. 2 April June

12 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis mistake to believe that we know from market observation what those discount rates should be. Capital markets are full of distortions related to the role of information. Further, market rates of return to investments are not social rates of return they generally take no account, for example, of environmental damages or other market distortions. But most importantly, we do not see any markets that can reveal clear answers to the question how do we, as a generation, value benefits to collective action to protect the climate for generations a hundred or more years from now? (see Hepburn, 2006, and Dietz et al., 2007, for further discussion). All this is not to argue that the markets contain no information. The problem is that they contain too much information on the one hand (real interest rates on 50-year, government-indexed bonds in the UK have been around half a percentage point recently and some investment rates of return are, say, 15% or 20% over the short or medium term) and too little information on the other (there are no substantive markets for the very long run). The interpretation of the data that are there needs very strong assumptions, which essentially force the answers. Notice too that governments vary greatly in what they impose as required rates of return on investments: for example, the UK government uses 3.5% for mediumterm projects (falling over the long term) and the US 7%. There are all sorts of institutional reasons for making these choices of required rates of return, including gaming and optimism bias from project sponsors. All this implies that for this type of non-marginal, very long-term issue, we must go back to first principles. With the restrictive assumption of marginal changes in the absence of uncertainty, the social discount rate r in these models is, where g is the growth rate: r = ηg + δ Each element on the right-hand side of (1) has a different role. First, η captures attitudes, in this framework, to risk and to intra-generational distribution, as well as to inter-generational distribution. Second, g is a feature of model structures and assumptions, not ethics. Higher g gives not only a higher social discount rate but also earlier emissions and hence earlier and higher damages from GHGs. Third, in the context of climate-change policy, δ is largely about ethical discrimination by date of birth (apart from the probability of planetary demise: see Chapter 2 of the Review). (1) 240 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

13 Right for the Right Reasons: A final rejoinder on the Stern Review Let us consider some of these elements in turn. First, we examine δ. We would still insist that we have not heard a serious ethical argument in favour of extreme values of δ of 2% or 3%, which Nordhaus (2006) and Weitzman (forthcoming) appear to support. Different values of δ will be appropriate in different circumstances. The circumstances here are collective choices today to reduce global emissions of GHGs, providing potentially very large benefits across many generations. Seen in this light, it is very clear that δ should largely be interpreted in terms of ethical discrimination by birth date. It is not a question of an individual s impatience with respect to his/her own consumption in his/her own lifetime, nor should it include the larger set of risks to the survival of individual government projects, with a marginal effect relative to the overall growth path. 11 Interpreted as discrimination by birth date, extreme values of δ are difficult to justify. If δ = 2%, then someone born in 1972 would have twice the ethical weight of someone born in 2007 (for δ = 1.5%, the date corresponding to 2007 is 2020). In other words, if they were both expected to have the same income, an extra unit of consumption to the latter would have half the weight to that of the former. Would many regard this as ethically acceptable in terms of responsible social action? We think not. Further, high δ can lead to a version of time inconsistency each generation postpones action, because with high δ it will also seek to minimise short-term mitigation costs, passing the burden on to the next generation, and so on. Beckerman and Hepburn (2007), drawing on the moral philosophy of David Hume, describe how agent-relative ethics may be justified by observing that such ethics have over the centuries protected us from anarchy and encouraged cooperation, first and foremost with those close to us. But, again, context is key: in climate change we face unprecedented risks to human development, where the causes and consequences of these risks are generations apart. Ultimately, strong international collective action can only be sustained by support from people around the world, but they must consider directly questions like how do we, as a generation, value benefits to collective action to protect the climate for generations a hundred or more years from now? For all the reasons we have given, it is simply mistaken to use market rates, or required rates on government projects, as an answer to questions of how people approach this issue. 11 Covered by, for example, the Green Book in the UK (HM Treasury, 2003). WORLD ECONOMICS Vol. 8 No. 2 April June

14 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis Next, we examine growth, g. The growth assumptions in the formal modelling of Chapter 6 of the Review were fairly conservative: global growth starts at around 2.5% on aggregate (0.9% per capita, due to rapid population growth) and falls to around 1.8% (1.4% per capita) in the latter half of the 22nd century. However, it is certainly plausible that over a period of time global growth rates could be higher than this. This would have two effects, working in opposite directions, on the assessment of future damages: first, higher growth brings both earlier emissions and thus damage; second, higher future incomes bring greater discounting (before the effects of climate damages kick in hard). We have not formally modelled these effects but our judgement and preliminary assessments suggest that both effects are strong. Let us now turn to η. Some have argued (e.g. Dasgupta, 2006) that η =1 is too low. This is also an ethical parameter and as such it is important to look at alternatives as we did in the Postscript to the Review. What is an appropriate range? Many cost benefit analyses use essentially η = 0: i.e. they weight an extra dollar to all individuals in the same way. That is problematic over an infinite horizon, since with growth it can lead to divergent welfare integrals (see the Appendix to Chapter 2 of the Review). On the other hand, η = 2, as some seem to propose, does not appear to be credible. Those who argue η = 2 are by implication saying that taking one dollar from an individual A, who has five times the income of individual B, is a social improvement, provided no more than 96% gets lost on the way (in other words, an extra dollar to individual B is worth 5 2 or 25 times that to individual A). That would be the ethical position implied and because η captures attitudes to inequality both within and between generations, consistency demands that we see such a redistribution as a social improvement not only between generations, but between individuals today. We have to go back to first principles in considering a range of values of η and δ. For η, we would suggest that the above discussion points to, for sensitivity analysis, a reasonable range of between 1 and 2, although we would suggest that the range 1 to 1.5 is likely to be of greater interest to most ethical observers. We do recognise that the combination of η = 1 and δ = 0.1% places a very high weight on the future (see the Appendix to Chapter 2 of the Review on convergence of utility integration). And we recognise that there is a plausible ethical case for higher η. It is a mistake, however, to argue that η = 1 together with low δ necessarily imply very 242 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

15 Right for the Right Reasons: A final rejoinder on the Stern Review high savings rates if incorporated into an optimum savings model (as Dasgupta, 2006, and Nordhaus, 2006, have done). The reason is that the optimum savings rates in such models also depend on assumptions about the structure of production, including technical progress. If, for example, technical progress contributes significantly to growth, then η = 1 together with low δ are consistent with current rates of savings. Overall, we would argue that, in focussing on discounting, our critics have shown little understanding of the great body of literature on applied welfare economics and project appraisal (a notable exception is Dasgupta, 2007). And they have taken little account of the very big differences between marginal and non-marginal analysis, between collective action and individual action, between the short to medium term and the long term (and the appropriate role of market data in that distinction), and between risk and uncertainty (although Weitzman, forthcoming, has much that is valuable on risk and uncertainty) Risk in formal economic modelling The structure of risks included in our formal modelling was cautious. In its Fourth Assessment Report, the IPCC is very careful not to include in its probability distributions of temperature increase the effects of phenomena that, although they appear to be real possibilities, have not yet been sufficiently modelled to provide relevant probabilities. Nevertheless it is recognised that such possibilities are there. Examples include abrupt, discontinuous and large-scale positive natural feedbacks that would amplify the warming caused by GHG emissions, such as the thawing of the permafrost to release the potent GHG methane, the collapse of the Amazon ecosystem and thus the loss of a major carbon sink, and the decline in the absorptive capacity of the oceans and other features of the carbon cycle. In the Review, we illustrated the effects of introducing a modest scenario of such positive feedbacks, based on studies that have explored methane releases from wetlands and melting permafrost, as well as weakened carbon sinks (see Box 6.1, on p. 175 of the Review). This modest scenario, which we called the high-climate scenario, added 0.4 C to mean warming in 2100 compared with the Review s baseline-climate scenario (which replicates the range of warming projections in the IPCC s Third Assessment Report), and an additional 1.2 C in The mean estimate of the total WORLD ECONOMICS Vol. 8 No. 2 April June

16 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis discounted cost of business-as-usual (BAU) climate change increased from around 11% to around 14%. 12 Other possible effects that have been identified in the scientific literature might exert still more powerful influences. Thus, there are strong grounds for thinking that the models of Chapter 6 of the Review underestimate risk. And we have seen that the effects of including just one of the relevant extra causes of risk may be high. The lines of argument on risk set out by Weitzman (forthcoming) are very interesting and indicate some agreement with this view Putting ethics and risk together In this paper, we have emphasised that the two fundamental issues guiding the appropriate strength and timing of climate-change policy are ethics and risk. The more hostile commentators on the Review s formal modelling usually adopt one of two positions. The first is that our basic conclusion is wrong. The second is that it is right, but for the wrong reasons. Those who believe our conclusion is wrong neglect, or adopt a misguided position on, one or both of ethics and risk. We discussed their objections above. Those who believe we are right but for the wrong reasons are merely choosing to emphasise one (usually risk) at the expense of the other (ethics). But both are necessary foundations of the case for strong action, as we argued in the Review and demonstrated in the Postscript and in Dietz et al. (2007). Figure 1 offers a different way to make the point. At the bottom of the diagram is the Review s estimate of the mean total discounted cost of BAU climate change, using the PAGE2002 integrated assessment model and our base modelling case. 13 In our base case, δ = 0.1%, η = 1 and we take risk into account by calculating expected utility from a wide range of scenarios, produced using Monte Carlo methods. The estimate is a permanent, 10.9% loss in global mean consumption per capita today. At the top of the diagram is an estimate that would follow from some previous studies and is implied by some critiques of our formal modelling 12 This measure of total discounted cost is derived from a comparison of the balanced growth equivalent or BGE of consumption without climate change to the BGE of consumption after climate damage and adaptation costs have been deducted. It summarises simulated losses over time, regions of the world and possible states of the world in terms of a permanent loss of global mean per-capita consumption today. 13 As is by now familiar in our sensitivity analyses, we consider the baseline-climate scenario, with market impacts, non-market impacts and the risk of abrupt, large-scale and discontinuous or catastrophic climatic changes. PAGE2002 is comprehensively reported in Hope (2006). 244 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

17 Right for the Right Reasons: A final rejoinder on the Stern Review Figure 1: The role of equity and risk in the Review s formal modelling A critic s case (δ = 1.5%, η = 2, no uncertainty): 0.6% Change ethics (δ = 0.1%, η = 1, no uncertainty): 3.5% Change uncertainty (δ = 1.5%, η = 2, expected-utility analysis): 1.1% Base Case (δ = 0.1%, η = 1, expected-utility analysis): 10.9% (e.g. Nordhaus, 2006): δ = 1.5% and η = 2, while little or no account is taken of risk or uncertainty, an effect we produce by taking mode values or best guesses for all the hitherto stochastic parameters in PAGE2002. The damage estimate is just 0.6%, too low to support strong action. In between, we show two very different routes between this implausible set of assumptions and our base case. Moving down the left-hand side, we decrease δ and η so as to place more ethical weight on future generations, but continue to omit uncertainty. The mean estimated cost of BAU climate change increases to 3.5%. Down the right-hand side, we retain the implausible ethical parameters of the critic s case, but take account of uncertainty by calculating expected utility, using the Monte Carlo methods and associated probability distributions of the Review. The cost of climate change increases only fractionally, to 1.1%. Neither change is alone sufficient to bridge the gap between the top and bottom examples. It is the interaction between risk and ethics that is crucial. This should be obvious: greater climate risks fall in the future, and it is only through affording future generations significant ethical weight that we would be motivated to protect them from these risks. Thus it is an error to suggest that our results, which give damages higher than most of the previous literature, come only from the different ethical parameters. They come, as we have insisted throughout the discussion, from a serious analysis of ethics, and WORLD ECONOMICS Vol. 8 No. 2 April June

18 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis from incorporating risk and analysis based on modern science. Much of the earlier economics literature has been remiss in its treatment of these key issues. What happens if we increase η, as for example Dasgupta (2006) has suggested, while at the same time placing more emphasis on the danger that climate change could inflict very high costs on growth and development, as for example Weitzman (forthcoming) has done? In Table 1, we show the sensitivity of the Review s estimates to the interaction between higher values of η and greater dangers from climate change. We make a simple representation of greater dangers through changes in the convexity of the relationship linking increasing temperatures with increasing damages in market and non-market sectors of the economy (the damage function exponent, γ). 14 In the Review, γ, like many other parameters, was sampled from a triangular probability distribution, in this case with a minimum value of 1 (damages are linear in temperature), a mode value of 1.3, and a maximum value of 3 (giving a mean value of 1.8). To demonstrate sensitivities, in this case we treat γ as deterministic and consider the range 2 to 3. γ = 2 corresponds most closely, in terms of estimated damages, to the distribution used in the base case in the Review: i.e. the top row of Table 1 gives results very similar to those in the Review. The sensitivity analysis in the Postscript to the Review provides other examples of ways of thinking about greater risks, by shifting the distribution of γ. Taking together the arguments for higher η (Dasgupta, 2006) and greater emphasis on risk (Weitzman, forthcoming), we have a move down Table 1: Sensitivity of total cost of climate change to damage function exponent and consumption elasticity of social marginal utility in baselineclimate scenario (mean BGE loss, 5% 95% confidence interval) Consumption elasticity of social marginal utility (η) Damage function exponent ( ) 6.0 ( ) 3.3 ( ) ( ) 10.0 ( ) 5.2 ( ) ( ) 29.3 ( ) 29.1 ( ) 14 We could also have investigated sensitivity to the parameters that jointly determine the probability of an abrupt and discontinuous climate catastrophe. The results are similar (see Dietz et al., forthcoming), but the greater number of parameters in the latter approach make for a more complicated exposition. 246 WORLD ECONOMICS Vol. 8 No. 2 April June 2007

19 Right for the Right Reasons: A final rejoinder on the Stern Review the diagonal of Table 1. η = 1.5 and γ = 2.5 is another plausible example, which could have served as a central case. In fact it gives a cost estimate of 10%, very close to that of the Review. The Review itself adopted as its central case ethical parameters that give a high weight to the future, but was conservative on risk. Taking more egalitarian values and allowing for the extra risk and a plausible case can be made for both gives another possible central theme with very similar results. We have stayed within the formal modelling and have shown that combining the position of some of the more thoughtful commentators gives results similar to those of the central case of the Review. We must emphasise very strongly, however, that the formal modelling leaves out key issues that would raise estimated damages beyond those embodied in the model. These were discussed explicitly and strongly both in the Review itself and in our earlier response to critiques in this journal (Dietz et al., 2007). Amongst them, we would emphasise: (i) the treatment of environmental goods and services as separate goods, in contrast to the aggregated treatment of climate damages in almost all studies, of which Chapter 6 of the Review was no exception; (ii) intra-generational distribution; and (iii) the risk of conflict. If incomes grow, but the environment is damaged due to BAU emissions, then the price of environmental goods, in terms of social willingness-to-pay, will rise sharply (see p. 58 of the Review). Thus investing in alternative (non-mitigation) investments with the intent of buying down climate damage later will very likely be a misguided policy, although it is implicit in the slow policy ramp of Nordhaus (2006) and others. This point has been convincingly elaborated in formal models, such as those by Guesnerie (2004), and Hoel and Sterner (2006). The point is still more forceful if one takes account of the irreversibilities in the flow-stock process of GHG accumulation and in the nature of damages. The formal modelling in the literature (ourselves included) has not focused strongly enough on intra-generational distribution, even though it will be the poor countries who are hit earliest and hardest. The Review in its disaggregated analysis stressed this very strongly (see Chapter 4, Chapter 13 and much of Parts V and VI). An important topic for those who want to spend more effort and resources on the formal modelling would be to include these issues more explicitly. This will raise damage estimates, particularly for higher η. WORLD ECONOMICS Vol. 8 No. 2 April June

20 S. Dietz, D. Anderson, N. Stern, C. Taylor & D. Zenghelis The third omission, conflict, receives no explicit attention in the formal modelling. Indeed, it is very difficult to capture this risk formally in a convincing way. However, the possibility of large temperature increases, say greater than 5 C above pre-industrial, would very likely involve massive pressures for movements of population away from the Equator. Remember that when temperatures were 5 C below where they are now, much of North America and Europe was under hundreds of metres of ice. When populations move on a large scale, there are risks of severe conflict. Such considerations have not been part of the formal modelling but are surely of great potential importance in the case of risks under BAU. 5. Conclusions It may be helpful for those who want to navigate through the maze of commentaries on the Stern Review to have a bullet-point summary of our reactions. The summary below concludes our rejoinder. 1. The costs of stabilising the stock of GHGs in the range ppm CO 2 e are considerably less than the costs of delayed action. This conclusion is robust across most reasonable perspectives, including parameter variation within formal modelling. 2. The policies proposed by the Review to stabilise within this range are sound and based on strong economic principles, which move beyond the previous literature in important ways, concerning risks and ethics and constructing an international deal. 3. The Review s foremost argument for strong action is based on a detailed, disaggregated assessment of the risks of business-as-usual (or of delayed action) in various regions and on various dimensions. The types and scale of risks involved were confirmed by the IPCC s Fourth Assessment Report a few months after publication of the Review, thus dismissing early claims by some that we exaggerated the risks, and by others that we understated them On the point of consistency between the IPCC and the Stern Review, it should be noted that the excerpt from a BBC radio interview, quoted by Henderson (2007, section 3: Parallel assessments ), is taken from a programme aired before the IPCC released its Fourth Assessment Report and therefore relates to a comparison with previous IPCC results, despite the way in which Henderson uses it (the programme was broadcast on 25 January 2007 and recorded weeks before that: IPCC released the Summary for Policy Makers of Working Group I on 2 February 2007). Furthermore, the statement by Stern was taken out of context and was not in response to the question put by the journalist, which was inserted afterwards. 248 WORLD ECONOMICS Vol. 8 No. 2 April June 2007